Why might hypertext not question the unity of the text deeply enough? I said this partly from experience. The hypertexts I've read so far don't do it as well as some more linear texts. Hypertext is good at questioning the borders of the text but there is more to unity than that.
You would think that hypertext would question authorship--George Landow certainly makes this claim--but so far I don't think that has happened much. Single author hypertexts, such as the Eastgate fictions, have their own virtues but so far they all have fairly strong implied authors. Multiple author hypertexts might show more promise in this regard but often they tend to just sprawl (see, for instance, the multi-author hypertext Grell that has too many authors, or the the more controlled, if that is the word, work in progress Scibe).
In multiple authored hypertexts there is also the problem of maintaining reader interest, since attention is a scarce resource. It may be that the reader learns to read for the momentary intensities of the text and its transitions, rather than for anything that might be called content or philosophy. I suppose that that might be considered the death of literacy. There are, however, more possibilities than a simple opposition of single intensities versus connected content. I am thinking not only of recent experimental writing (in linear books) but also of older kinds of multiple authorship, in particular the Japanese kasen (a kind of renga or linked verse), multiply authored poem sequences within a complex set of conventions, with parts that are seemingly unrelated at first glance, then showing a kind of unity that doesn't fit into our standard oppositions.